Cross-situational word learning model, learning the meanings of words from image features.

First you will need to download the image vectors from the image-caption data sets and their corresponding captions. These are made available on http://cs.stanford.edu/people/karpathy/deepimagesent/ and detailed instructions can be found in the /dataset/ subfolder of this repo.

To create the visual word-vectors from the publication [coming up] from the Flickr8k data set you need to run

python train.py vector online f8k train test